socket.c

Go to the documentation of this file.

/*
 *  OpenVPN -- An application to securely tunnel IP networks
 *             over a single TCP/UDP port, with support for SSL/TLS-based
 *             session authentication and key exchange,
 *             packet encryption, packet authentication, and
 *             packet compression.
 *
 *  Copyright (C) 2002-2026 OpenVPN Inc <sales@openvpn.net>
 *
 *  This program is free software; you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License version 2
 *  as published by the Free Software Foundation.
 *
 *  This program is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License along
 *  with this program; if not, see <https://www.gnu.org/licenses/>.
 */
 
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
 
#include "syshead.h"
 
#include "socket.h"
#include "fdmisc.h"
#include "misc.h"
#include "gremlin.h"
#include "plugin.h"
#include "ps.h"
#include "run_command.h"
#include "manage.h"
#include "misc.h"
#include "manage.h"
#include "openvpn.h"
#include "forward.h"
#include "resolve.h"
 
#include "memdbg.h"
 
bool
sockets_read_residual(const struct context *c)
{
    for (int i = 0; i < c->c1.link_sockets_num; i++)
    {
        if (c->c2.link_sockets[i]->stream_buf.residual_fully_formed)
        {
            return true;
        }
    }
    return false;
}
 
static int
socket_get_sndbuf(socket_descriptor_t sd)
{
    int val;
    socklen_t len = sizeof(val);
 
    if (getsockopt(sd, SOL_SOCKET, SO_SNDBUF, (void *)&val, &len) == 0 && len == sizeof(val))
    {
        return val;
    }
    return 0;
}
 
static void
socket_set_sndbuf(socket_descriptor_t sd, int size)
{
    if (setsockopt(sd, SOL_SOCKET, SO_SNDBUF, (void *)&size, sizeof(size)) != 0)
    {
        msg(M_WARN, "NOTE: setsockopt SO_SNDBUF=%d failed", size);
    }
}
 
static int
socket_get_rcvbuf(socket_descriptor_t sd)
{
    int val;
    socklen_t len = sizeof(val);
 
    if (getsockopt(sd, SOL_SOCKET, SO_RCVBUF, (void *)&val, &len) == 0 && len == sizeof(val))
    {
        return val;
    }
    return 0;
}
 
static void
socket_set_rcvbuf(socket_descriptor_t sd, int size)
{
    if (setsockopt(sd, SOL_SOCKET, SO_RCVBUF, (void *)&size, sizeof(size)) != 0)
    {
        msg(M_WARN, "NOTE: setsockopt SO_RCVBUF=%d failed", size);
    }
}
 
void
socket_set_buffers(socket_descriptor_t fd, const struct socket_buffer_size *sbs, bool reduce_size)
{
    if (sbs)
    {
        const int sndbuf_old = socket_get_sndbuf(fd);
        const int rcvbuf_old = socket_get_rcvbuf(fd);
 
        if (sbs->sndbuf && (reduce_size || sndbuf_old < sbs->sndbuf))
        {
            socket_set_sndbuf(fd, sbs->sndbuf);
        }
 
        if (sbs->rcvbuf && (reduce_size || rcvbuf_old < sbs->rcvbuf))
        {
            socket_set_rcvbuf(fd, sbs->rcvbuf);
        }
 
        msg(D_OSBUF, "Socket Buffers: R=[%d->%d] S=[%d->%d]", rcvbuf_old, socket_get_rcvbuf(fd),
            sndbuf_old, socket_get_sndbuf(fd));
    }
}
 
/*
 * Set other socket options
 */
 
static bool
socket_set_tcp_nodelay(socket_descriptor_t sd, int state)
{
#if defined(_WIN32) || (defined(IPPROTO_TCP) && defined(TCP_NODELAY))
    if (setsockopt(sd, IPPROTO_TCP, TCP_NODELAY, (void *)&state, sizeof(state)) != 0)
    {
        msg(M_WARN, "NOTE: setsockopt TCP_NODELAY=%d failed", state);
        return false;
    }
    else
    {
        dmsg(D_OSBUF, "Socket flags: TCP_NODELAY=%d succeeded", state);
        return true;
    }
#else /* if defined(_WIN32) || (defined(IPPROTO_TCP) && defined(TCP_NODELAY)) */
    msg(M_WARN, "NOTE: setsockopt TCP_NODELAY=%d failed (No kernel support)", state);
    return false;
#endif
}
 
static inline void
socket_set_mark(socket_descriptor_t sd, int mark)
{
#if defined(TARGET_LINUX)
    if (mark && setsockopt(sd, SOL_SOCKET, SO_MARK, (void *)&mark, sizeof(mark)) != 0)
    {
        msg(M_WARN, "NOTE: setsockopt SO_MARK=%d failed", mark);
    }
#endif
}
 
void
link_socket_update_buffer_sizes(struct link_socket *sock, int rcvbuf, int sndbuf)
{
    if (sock && socket_defined(sock->sd))
    {
        sock->socket_buffer_sizes.sndbuf = sndbuf;
        sock->socket_buffer_sizes.rcvbuf = rcvbuf;
        socket_set_buffers(sock->sd, &sock->socket_buffer_sizes, true);
    }
}
 
/*
 * SOCKET INITIALIZATION CODE.
 * Create a TCP/UDP socket
 */
 
socket_descriptor_t
create_socket_tcp(struct addrinfo *addrinfo)
{
    socket_descriptor_t sd;
 
    ASSERT(addrinfo);
    ASSERT(addrinfo->ai_socktype == SOCK_STREAM);
 
    if ((sd = socket(addrinfo->ai_family, addrinfo->ai_socktype, addrinfo->ai_protocol))
        == SOCKET_UNDEFINED)
    {
        msg(M_ERR, "Cannot create TCP socket");
    }
 
#ifndef _WIN32 /* using SO_REUSEADDR on Windows will cause bind to succeed on port conflicts! */
    /* set SO_REUSEADDR on socket */
    {
        int on = 1;
        if (setsockopt(sd, SOL_SOCKET, SO_REUSEADDR, (void *)&on, sizeof(on)) < 0)
        {
            msg(M_ERR, "TCP: Cannot setsockopt SO_REUSEADDR on TCP socket");
        }
    }
#endif
 
    /* set socket file descriptor to not pass across execs, so that
     * scripts don't have access to it */
    set_cloexec(sd);
 
    return sd;
}
 
static socket_descriptor_t
create_socket_udp(struct addrinfo *addrinfo, const unsigned int flags, bool optional)
{
    socket_descriptor_t sd;
 
    ASSERT(addrinfo);
    ASSERT(addrinfo->ai_socktype == SOCK_DGRAM);
 
    if ((sd = socket(addrinfo->ai_family, addrinfo->ai_socktype, addrinfo->ai_protocol))
        == SOCKET_UNDEFINED)
    {
        msg(optional ? D_LOW | M_ERRNO : M_ERR, "UDP: Cannot create UDP/UDP6 socket");
        return SOCKET_UNDEFINED;
    }
#if ENABLE_IP_PKTINFO
    else if (flags & SF_USE_IP_PKTINFO)
    {
        int pad = 1;
        if (addrinfo->ai_family == AF_INET)
        {
#if defined(HAVE_IN_PKTINFO) && defined(HAVE_IPI_SPEC_DST)
            if (setsockopt(sd, SOL_IP, IP_PKTINFO, (void *)&pad, sizeof(pad)) < 0)
            {
                msg(M_ERR, "UDP: failed setsockopt for IP_PKTINFO");
            }
#elif defined(IP_RECVDSTADDR)
            if (setsockopt(sd, IPPROTO_IP, IP_RECVDSTADDR, (void *)&pad, sizeof(pad)) < 0)
            {
                msg(M_ERR, "UDP: failed setsockopt for IP_RECVDSTADDR");
            }
#else /* if defined(HAVE_IN_PKTINFO) && defined(HAVE_IPI_SPEC_DST) */
#error ENABLE_IP_PKTINFO is set without IP_PKTINFO xor IP_RECVDSTADDR (fix syshead.h)
#endif
        }
        else if (addrinfo->ai_family == AF_INET6)
        {
#ifndef IPV6_RECVPKTINFO /* Some older Darwin platforms require this */
            if (setsockopt(sd, IPPROTO_IPV6, IPV6_PKTINFO, (void *)&pad, sizeof(pad)) < 0)
#else
            if (setsockopt(sd, IPPROTO_IPV6, IPV6_RECVPKTINFO, (void *)&pad, sizeof(pad)) < 0)
#endif
            {
                msg(M_ERR, "UDP: failed setsockopt for IPV6_RECVPKTINFO");
            }
        }
    }
#endif /* if ENABLE_IP_PKTINFO */
 
    /* set socket file descriptor to not pass across execs, so that
     * scripts don't have access to it */
    set_cloexec(sd);
 
    return sd;
}
 
static void
bind_local(const struct link_socket *sock)
{
    /* bind to local address/port */
    if (sock->bind_local)
    {
        if (sock->socks_proxy && sock->info.proto == PROTO_UDP)
        {
            socket_bind(sock->ctrl_sd, sock->info.lsa->bind_local, sock->info.af, "SOCKS", false);
        }
        else
        {
            socket_bind(sock->sd, sock->info.lsa->bind_local, sock->info.af, "TCP/UDP",
                        sock->info.bind_ipv6_only);
        }
    }
}
 
/* The per-socket options every link socket gets right after creation. */
static void
socket_apply_options(socket_descriptor_t sd, const struct socket_buffer_size *sbs, int mark,
                     const char *bind_dev)
{
    /* set socket buffers based on --sndbuf and --rcvbuf options */
    socket_set_buffers(sd, sbs, true);
 
    /* set socket to --mark packets with given value */
    socket_set_mark(sd, mark);
 
#if defined(TARGET_LINUX)
    if (bind_dev)
    {
        msg(M_INFO, "Using bind-dev %s", bind_dev);
        /* Note: We verify strlen of bind_dev in options parsing */
        if (setsockopt(sd, SOL_SOCKET, SO_BINDTODEVICE, bind_dev, (socklen_t)(strlen(bind_dev) + 1))
            != 0)
        {
            msg(M_WARN | M_ERRNO, "WARN: setsockopt SO_BINDTODEVICE=%s failed", bind_dev);
        }
    }
#else
    (void)bind_dev;
#endif
}
 
socket_descriptor_t
create_socket_udp_configured(sa_family_t af, unsigned int sockflags, const struct socket_buffer_size *sbs,
                             int mark, const char *bind_dev, struct addrinfo *bind_addr,
                             bool bind_ipv6_only, bool optional)
{
    struct addrinfo ai = { .ai_family = af, .ai_socktype = SOCK_DGRAM, .ai_protocol = IPPROTO_UDP };
    socket_descriptor_t sd = create_socket_udp(&ai, sockflags, optional);
 
    if (sd == SOCKET_UNDEFINED)
    {
        return SOCKET_UNDEFINED;
    }
 
    socket_apply_options(sd, sbs, mark, bind_dev);
    if (bind_addr)
    {
        socket_bind(sd, bind_addr, af, "TCP/UDP", bind_ipv6_only);
    }
    return sd;
}
 
static void
create_socket(struct link_socket *sock, struct addrinfo *addr)
{
    /* Set af field of sock->info, so it always reflects the address family
     * of the created socket */
    sock->info.af = (sa_family_t)addr->ai_family;
 
    if (addr->ai_protocol == IPPROTO_UDP || addr->ai_socktype == SOCK_DGRAM)
    {
        /* With a SOCKS proxy the local bind goes on the control socket instead
         * (see bind_local()), so the UDP socket is created unbound then. */
        struct addrinfo *bind_addr =
            (sock->bind_local && !sock->socks_proxy) ? sock->info.lsa->bind_local : NULL;
        sock->sd = create_socket_udp_configured(sock->info.af, sock->sockflags, &sock->socket_buffer_sizes,
                                                sock->mark, sock->bind_dev, bind_addr,
                                                sock->info.bind_ipv6_only, false);
        sock->sockflags |= SF_GETADDRINFO_DGRAM;
 
        /* Assume that control socket and data socket to the socks proxy
         * are using the same IP family */
        if (sock->socks_proxy)
        {
            /* Construct a temporary addrinfo to create the socket,
             * currently resolve two remote addresses is not supported,
             * TODO: Rewrite the whole resolve_remote */
            struct addrinfo addrinfo_tmp = *addr;
            addrinfo_tmp.ai_socktype = SOCK_STREAM;
            addrinfo_tmp.ai_protocol = IPPROTO_TCP;
            sock->ctrl_sd = create_socket_tcp(&addrinfo_tmp);
            bind_local(sock);
        }
    }
    else if (addr->ai_protocol == IPPROTO_TCP || addr->ai_socktype == SOCK_STREAM)
    {
        sock->sd = create_socket_tcp(addr);
        socket_apply_options(sock->sd, &sock->socket_buffer_sizes, sock->mark, sock->bind_dev);
        bind_local(sock);
    }
    else
    {
        ASSERT(0);
    }
}
 
#ifdef TARGET_ANDROID
static void
protect_fd_nonlocal(int fd, const struct sockaddr *addr)
{
    if (!management)
    {
        msg(M_FATAL, "Required management interface not available.");
    }
 
    /* pass socket FD to management interface to pass on to VPNService API
     * as "protected socket" (exempt from being routed into tunnel)
     */
    if (addr_local(addr))
    {
        msg(D_SOCKET_DEBUG, "Address is local, not protecting socket fd %d", fd);
        return;
    }
 
    msg(D_SOCKET_DEBUG, "Protecting socket fd %d", fd);
    management->connection.fdtosend = fd;
    management_android_control(management, "PROTECTFD", __func__);
}
#endif
 
/*
 * Functions used for establishing a TCP stream connection.
 */
static void
socket_do_listen(socket_descriptor_t sd, const struct addrinfo *local, bool do_listen,
                 bool do_set_nonblock)
{
    struct gc_arena gc = gc_new();
    if (do_listen)
    {
        ASSERT(local);
        msg(M_INFO, "Listening for incoming TCP connection on %s",
            print_sockaddr(local->ai_addr, &gc));
        if (listen(sd, 32))
        {
            msg(M_ERR, "TCP: listen() failed");
        }
    }
 
    /* set socket to non-blocking mode */
    if (do_set_nonblock)
    {
        set_nonblock(sd);
    }
 
    gc_free(&gc);
}
 
#if defined(__GNUC__) || defined(__clang__)
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wsign-conversion"
#endif
 
socket_descriptor_t
socket_do_accept(socket_descriptor_t sd, struct link_socket_actual *act, const bool nowait)
{
    /* af_addr_size WILL return 0 in this case if AFs other than AF_INET
     * are compiled because act is empty here.
     * could use getsockname() to support later remote_len check
     */
    socklen_t remote_len_af = af_addr_size(act->dest.addr.sa.sa_family);
    socklen_t remote_len = sizeof(act->dest.addr);
    socket_descriptor_t new_sd;
 
    CLEAR(*act);
 
    if (nowait)
    {
        new_sd = getpeername(sd, &act->dest.addr.sa, &remote_len);
 
        if (!socket_defined(new_sd))
        {
            msg(D_LINK_ERRORS | M_ERRNO, "TCP: getpeername() failed");
        }
        else
        {
            new_sd = sd;
        }
    }
    else
    {
        new_sd = accept(sd, &act->dest.addr.sa, &remote_len);
    }
 
#if 0 /* For debugging only, test the effect of accept() failures */
    {
        static int foo = 0;
        ++foo;
        if (foo & 1)
        {
            new_sd = -1;
        }
    }
#endif
 
    if (!socket_defined(new_sd))
    {
        msg(D_LINK_ERRORS | M_ERRNO, "TCP: accept(%d) failed", (int)sd);
    }
    /* only valid if we have remote_len_af!=0 */
    else if (remote_len_af && remote_len != remote_len_af)
    {
        msg(D_LINK_ERRORS,
            "TCP: Received strange incoming connection with unknown address length=%u",
            (unsigned int)remote_len);
        openvpn_close_socket(new_sd);
        new_sd = SOCKET_UNDEFINED;
    }
    else
    {
        /* set socket file descriptor to not pass across execs, so that
         * scripts don't have access to it */
        set_cloexec(new_sd);
    }
    return new_sd;
}
 
static void
tcp_connection_established(const struct link_socket_actual *act)
{
    struct gc_arena gc = gc_new();
    msg(M_INFO, "TCP connection established with %s", print_link_socket_actual(act, &gc));
    gc_free(&gc);
}
 
static socket_descriptor_t
socket_listen_accept(socket_descriptor_t sd, struct link_socket_actual *act,
                     const struct addrinfo *local, bool do_listen,
                     bool nowait, volatile int *signal_received)
{
    socket_descriptor_t new_sd = SOCKET_UNDEFINED;
 
    CLEAR(*act);
    socket_do_listen(sd, local, do_listen, true);
 
    while (true)
    {
        int status;
        fd_set reads;
        struct timeval tv;
 
        FD_ZERO(&reads);
        openvpn_fd_set(sd, &reads);
        tv.tv_sec = 0;
        tv.tv_usec = 0;
 
        status = openvpn_select(sd + 1, &reads, NULL, NULL, &tv);
 
        get_signal(signal_received);
        if (*signal_received)
        {
            return sd;
        }
 
        if (status < 0)
        {
            msg(D_LINK_ERRORS | M_ERRNO, "TCP: select() failed");
        }
 
        if (status <= 0)
        {
            management_sleep(1);
            continue;
        }
 
        new_sd = socket_do_accept(sd, act, nowait);
 
        if (socket_defined(new_sd))
        {
            break;
        }
        management_sleep(1);
    }
 
    if (!nowait && openvpn_close_socket(sd))
    {
        msg(M_ERR, "TCP: close socket failed (sd)");
    }
 
    tcp_connection_established(act);
 
    return new_sd;
}
 
void
socket_bind(socket_descriptor_t sd, const struct addrinfo *local, int ai_family,
            const char *prefix, bool ipv6only)
{
    struct gc_arena gc = gc_new();
 
    /* FIXME (schwabe)
     * getaddrinfo for the bind address might return multiple AF_INET/AF_INET6
     * entries for the requested protocol.
     * For example if an address has multiple A records
     * What is the correct way to deal with it?
     */
 
    ASSERT(local);
 
    /* find the first addrinfo with correct ai_family */
    const struct addrinfo *cur;
    for (cur = local; cur; cur = cur->ai_next)
    {
        if (cur->ai_family == ai_family)
        {
            break;
        }
    }
    if (!cur)
    {
        msg(M_FATAL, "%s: Socket bind failed: Addr to bind has no %s record", prefix,
            addr_family_name(ai_family));
    }
 
    if (ai_family == AF_INET6)
    {
        int v6only = ipv6only ? 1 : 0; /* setsockopt must have an "int" */
 
        msg(M_INFO, "setsockopt(IPV6_V6ONLY=%d)", v6only);
        if (setsockopt(sd, IPPROTO_IPV6, IPV6_V6ONLY, (void *)&v6only, sizeof(v6only)))
        {
            msg(M_NONFATAL | M_ERRNO, "Setting IPV6_V6ONLY=%d failed", v6only);
        }
    }
    if (openvpn_bind(sd, cur->ai_addr, cur->ai_addrlen))
    {
        msg(M_FATAL | M_ERRNO, "%s: Socket bind failed on local address %s", prefix,
            print_sockaddr_ex(local->ai_addr, ":", PS_SHOW_PORT, &gc));
    }
    gc_free(&gc);
}
 
int
openvpn_connect(socket_descriptor_t sd, const struct sockaddr *remote, int connect_timeout,
                volatile int *signal_received)
{
    int status = 0;
 
#ifdef TARGET_ANDROID
    protect_fd_nonlocal(sd, remote);
#endif
    set_nonblock(sd);
    status = connect(sd, remote, af_addr_size(remote->sa_family));
    if (status)
    {
        status = openvpn_errno();
    }
    if (
#ifdef _WIN32
        status == WSAEWOULDBLOCK
#else
        status == EINPROGRESS
#endif
    )
    {
        while (true)
        {
#if POLL
            struct pollfd fds[1];
            fds[0].fd = sd;
            fds[0].events = POLLOUT;
            status = poll(fds, 1, (connect_timeout > 0) ? 1000 : 0);
#else
            fd_set writes;
            struct timeval tv;
 
            FD_ZERO(&writes);
            openvpn_fd_set(sd, &writes);
            tv.tv_sec = (connect_timeout > 0) ? 1 : 0;
            tv.tv_usec = 0;
 
            status = openvpn_select(sd + 1, NULL, &writes, NULL, &tv);
#endif
            if (signal_received)
            {
                get_signal(signal_received);
                if (*signal_received)
                {
                    status = 0;
                    break;
                }
            }
            if (status < 0)
            {
                status = openvpn_errno();
                break;
            }
            if (status <= 0)
            {
                if (--connect_timeout < 0)
                {
#ifdef _WIN32
                    status = WSAETIMEDOUT;
#else
                    status = ETIMEDOUT;
#endif
                    break;
                }
                management_sleep(0);
                continue;
            }
 
            /* got it */
            {
                int val = 0;
                socklen_t len;
 
                len = sizeof(val);
                if (getsockopt(sd, SOL_SOCKET, SO_ERROR, (void *)&val, &len) == 0
                    && len == sizeof(val))
                {
                    status = val;
                }
                else
                {
                    status = openvpn_errno();
                }
                break;
            }
        }
    }
 
    return status;
}
 
void
set_actual_address(struct link_socket_actual *actual, struct addrinfo *ai)
{
    CLEAR(*actual);
    ASSERT(ai);
 
    if (ai->ai_family == AF_INET)
    {
        actual->dest.addr.in4 = *((struct sockaddr_in *)ai->ai_addr);
    }
    else if (ai->ai_family == AF_INET6)
    {
        actual->dest.addr.in6 = *((struct sockaddr_in6 *)ai->ai_addr);
    }
    else
    {
        ASSERT(0);
    }
}
 
static void
socket_connect(socket_descriptor_t *sd, const struct sockaddr *dest, const int connect_timeout,
               struct signal_info *sig_info)
{
    struct gc_arena gc = gc_new();
    int status;
 
    msg(M_INFO, "Attempting to establish TCP connection with %s", print_sockaddr(dest, &gc));
 
#ifdef ENABLE_MANAGEMENT
    if (management)
    {
        management_set_state(management, OPENVPN_STATE_TCP_CONNECT, NULL, NULL, NULL, NULL, NULL);
    }
#endif
 
    /* Set the actual address */
    status = openvpn_connect(*sd, dest, connect_timeout, &sig_info->signal_received);
 
    get_signal(&sig_info->signal_received);
    if (sig_info->signal_received)
    {
        goto done;
    }
 
    if (status)
    {
        msg(D_LINK_ERRORS, "TCP: connect to %s failed: %s", print_sockaddr(dest, &gc),
            strerror(status));
 
        openvpn_close_socket(*sd);
        *sd = SOCKET_UNDEFINED;
        register_signal(sig_info, SIGUSR1, "connection-failed");
    }
    else
    {
        msg(M_INFO, "TCP connection established with %s", print_sockaddr(dest, &gc));
    }
 
done:
    gc_free(&gc);
}
 
/*
 * Stream buffer handling prototypes -- stream_buf is a helper class
 * to assist in the packetization of stream transport protocols
 * such as TCP.
 */
 
static void stream_buf_init(struct stream_buf *sb, const struct buffer *buf,
                            const unsigned int sockflags, const int proto);
 
static void stream_buf_close(struct stream_buf *sb);
 
static bool stream_buf_added(struct stream_buf *sb, ssize_t length_added);
 
/* For stream protocols, allocate a buffer to build up packet.
 * Called after frame has been finalized. */
 
static void
socket_frame_init(const struct frame *frame, struct link_socket *sock)
{
#ifdef _WIN32
    overlapped_io_init(&sock->reads, frame, FALSE);
    overlapped_io_init(&sock->writes, frame, TRUE);
    sock->rw_handle.read = sock->reads.overlapped.hEvent;
    sock->rw_handle.write = sock->writes.overlapped.hEvent;
#endif
 
    if (link_socket_connection_oriented(sock))
    {
#ifdef _WIN32
        stream_buf_init(&sock->stream_buf, &sock->reads.buf_init, sock->sockflags,
                        sock->info.proto);
#else
        alloc_buf_sock_tun(&sock->stream_buf_data, frame);
 
        stream_buf_init(&sock->stream_buf, &sock->stream_buf_data, sock->sockflags,
                        sock->info.proto);
#endif
    }
}
 
static void
resolve_bind_local(struct link_socket *sock)
{
    /* resolve local address if undefined */
    if (!sock->info.lsa->bind_local)
    {
        unsigned int flags = GETADDR_RESOLVE | GETADDR_WARN_ON_SIGNAL | GETADDR_FATAL | GETADDR_PASSIVE;
        int status;
 
        if (proto_is_dgram(sock->info.proto))
        {
            flags |= GETADDR_DATAGRAM;
        }
 
        /* will return AF_{INET|INET6}from local_host */
        status = get_cached_dns_entry(sock->dns_cache, sock->local_host, sock->local_port, sock->info.af,
                                      flags, &sock->info.lsa->bind_local);
 
        if (status)
        {
            status = openvpn_getaddrinfo(flags, sock->local_host, sock->local_port, 0, NULL, sock->info.af,
                                         &sock->info.lsa->bind_local);
        }
 
        if (status != 0)
        {
            msg(M_FATAL, "getaddrinfo() failed for local \"%s:%s\": %s", sock->local_host,
                sock->local_port, gai_strerror(status));
        }
 
        /* the address family returned by openvpn_getaddrinfo() should be
         * taken into consideration only if we really passed an hostname
         * to resolve. Otherwise its value is not useful to us and may
         * actually break our socket, i.e. when it returns AF_INET
         * but our remote is v6 only.
         */
        if (sock->local_host)
        {
            /* the resolved 'local entry' might have a different family than
             * what was globally configured
             */
            sock->info.af = (sa_family_t)sock->info.lsa->bind_local->ai_family;
        }
    }
}
 
static void
resolve_remote(struct link_socket *sock, int phase, struct signal_info *sig_info)
{
    volatile int *signal_received = sig_info ? &sig_info->signal_received : NULL;
    struct gc_arena gc = gc_new();
 
    /* resolve remote address if undefined */
    if (!sock->info.lsa->remote_list)
    {
        if (sock->remote_host)
        {
            unsigned int flags =
                sf2gaf(GETADDR_RESOLVE | GETADDR_UPDATE_MANAGEMENT_STATE, sock->sockflags);
            int retry = 0;
            int status = -1;
            struct addrinfo *ai;
            if (proto_is_dgram(sock->info.proto))
            {
                flags |= GETADDR_DATAGRAM;
            }
 
            if (sock->resolve_retry_seconds == RESOLV_RETRY_INFINITE)
            {
                if (phase == 2)
                {
                    flags |= (GETADDR_TRY_ONCE | GETADDR_FATAL);
                }
                retry = 0;
            }
            else if (phase == 1)
            {
                if (sock->resolve_retry_seconds)
                {
                    retry = 0;
                }
                else
                {
                    flags |= (GETADDR_FATAL | GETADDR_MENTION_RESOLVE_RETRY);
                    retry = 0;
                }
            }
            else if (phase == 2)
            {
                if (sock->resolve_retry_seconds)
                {
                    flags |= GETADDR_FATAL;
                    retry = sock->resolve_retry_seconds;
                }
                else
                {
                    ASSERT(0);
                }
            }
            else
            {
                ASSERT(0);
            }
 
 
            status = get_cached_dns_entry(sock->dns_cache, sock->remote_host, sock->remote_port,
                                          sock->info.af, flags, &ai);
            if (status)
            {
                status = openvpn_getaddrinfo(flags, sock->remote_host, sock->remote_port, retry,
                                             sig_info, sock->info.af, &ai);
            }
 
            if (status == 0)
            {
                sock->info.lsa->remote_list = ai;
                sock->info.lsa->current_remote = ai;
 
                dmsg(D_SOCKET_DEBUG, "RESOLVE_REMOTE flags=0x%04x phase=%d rrs=%d sig=%d status=%d",
                     flags, phase, retry, signal_received ? *signal_received : -1, status);
            }
            if (signal_received && *signal_received)
            {
                goto done;
            }
            if (status != 0)
            {
                if (signal_received)
                {
                    /* potential overwrite of signal */
                    register_signal(sig_info, SIGUSR1, "socks-resolve-failure");
                }
                goto done;
            }
        }
    }
 
    /* should we re-use previous active remote address? */
    if (link_socket_actual_defined(&sock->info.lsa->actual))
    {
        msg(M_INFO, "TCP/UDP: Preserving recently used remote address: %s",
            print_link_socket_actual(&sock->info.lsa->actual, &gc));
    }
    else
    {
        CLEAR(sock->info.lsa->actual);
        if (sock->info.lsa->current_remote)
        {
            set_actual_address(&sock->info.lsa->actual, sock->info.lsa->current_remote);
        }
    }
 
done:
    gc_free(&gc);
}
 
 
struct link_socket *
link_socket_new(void)
{
    struct link_socket *sock;
 
    ALLOC_OBJ_CLEAR(sock, struct link_socket);
    sock->sd = SOCKET_UNDEFINED;
    sock->ctrl_sd = SOCKET_UNDEFINED;
    sock->ev_arg.type = EVENT_ARG_LINK_SOCKET;
    sock->ev_arg.u.sock = sock;
 
    return sock;
}
 
void
link_socket_init_phase1(struct context *c, int sock_index, int mode)
{
    struct link_socket *sock = c->c2.link_sockets[sock_index];
    struct options *o = &c->options;
    ASSERT(sock);
 
    const char *host = o->ce.local_list->array[sock_index]->local;
    const char *port = o->ce.local_list->array[sock_index]->port;
    int proto = o->ce.local_list->array[sock_index]->proto;
    const char *remote_host = o->ce.remote;
    const char *remote_port = o->ce.remote_port;
 
    if (remote_host)
    {
        proto = o->ce.proto;
    }
 
    /* If --lport is specified in a client connection block,
     * it takes precedence over the global setting. */
    if (o->mode == MODE_POINT_TO_POINT && o->ce.local_port_defined)
    {
        port = o->ce.local_port;
    }
 
    if (c->mode == CM_CHILD_TCP || c->mode == CM_CHILD_UDP)
    {
        struct link_socket *tmp_sock = NULL;
        if (c->mode == CM_CHILD_TCP)
        {
            tmp_sock = (struct link_socket *)c->c2.accept_from;
        }
        else if (c->mode == CM_CHILD_UDP)
        {
            tmp_sock = c->c2.link_sockets[0];
        }
 
        host = tmp_sock->local_host;
        port = tmp_sock->local_port;
        proto = tmp_sock->info.proto;
    }
 
    sock->local_host = host;
    sock->local_port = port;
    sock->remote_host = remote_host;
    sock->remote_port = remote_port;
    sock->dns_cache = c->c1.dns_cache;
    sock->http_proxy = c->c1.http_proxy;
    sock->socks_proxy = c->c1.socks_proxy;
    sock->bind_local = o->ce.bind_local;
    sock->resolve_retry_seconds = o->resolve_retry_seconds;
    sock->mtu_discover_type = o->ce.mtu_discover_type;
 
#ifdef ENABLE_DEBUG
    sock->gremlin = o->gremlin;
#endif
 
    sock->socket_buffer_sizes.rcvbuf = o->rcvbuf;
    sock->socket_buffer_sizes.sndbuf = o->sndbuf;
 
    sock->sockflags = o->sockflags;
 
#if PORT_SHARE
    if (o->port_share_host && o->port_share_port)
    {
        sock->sockflags |= SF_PORT_SHARE;
    }
#endif
 
    sock->mark = o->mark;
    sock->bind_dev = o->bind_dev;
    ASSERT(proto >= 0 && proto < PROTO_N);
    sock->info.proto = (uint8_t)proto;
    sock->info.af = o->ce.af;
    sock->info.remote_float = o->ce.remote_float;
    sock->info.lsa = &c->c1.link_socket_addrs[sock_index];
    sock->info.bind_ipv6_only = o->ce.bind_ipv6_only;
    sock->info.ipchange_command = o->ipchange;
    sock->info.plugins = c->plugins;
    sock->server_poll_timeout = &c->c2.server_poll_interval;
 
    sock->mode = mode;
    if (mode == LS_MODE_TCP_ACCEPT_FROM)
    {
        ASSERT(c->c2.accept_from);
        ASSERT(sock->info.proto == PROTO_TCP_SERVER);
        sock->sd = c->c2.accept_from->sd;
        /* inherit (possibly guessed) info AF from parent context */
        sock->info.af = c->c2.accept_from->info.af;
    }
 
    /* are we running in HTTP proxy mode? */
    if (sock->http_proxy)
    {
        ASSERT(sock->info.proto == PROTO_TCP_CLIENT);
 
        /* the proxy server */
        sock->remote_host = c->c1.http_proxy->options.server;
        sock->remote_port = c->c1.http_proxy->options.port;
 
        /* the OpenVPN server we will use the proxy to connect to */
        sock->proxy_dest_host = remote_host;
        sock->proxy_dest_port = remote_port;
    }
    /* or in Socks proxy mode? */
    else if (sock->socks_proxy)
    {
        /* the proxy server */
        sock->remote_host = c->c1.socks_proxy->server;
        sock->remote_port = c->c1.socks_proxy->port;
 
        /* the OpenVPN server we will use the proxy to connect to */
        sock->proxy_dest_host = remote_host;
        sock->proxy_dest_port = remote_port;
    }
    else
    {
        sock->remote_host = remote_host;
        sock->remote_port = remote_port;
    }
 
    /* bind behavior for TCP server vs. client */
    if (sock->info.proto == PROTO_TCP_SERVER)
    {
        if (sock->mode == LS_MODE_TCP_ACCEPT_FROM)
        {
            sock->bind_local = false;
        }
        else
        {
            sock->bind_local = true;
        }
    }
 
    if (mode != LS_MODE_TCP_ACCEPT_FROM)
    {
        if (sock->bind_local)
        {
            resolve_bind_local(sock);
        }
        resolve_remote(sock, 1, NULL);
    }
}
 
static void
phase2_set_socket_flags(const struct link_socket *sock)
{
    /* TCP_NODELAY is enabled by default on every TCP socket; dco-win is
     * skipped as it manages its own socket */
    if (proto_is_tcp(sock->info.proto) && !(sock->sockflags & SF_DCO_WIN))
    {
        socket_set_tcp_nodelay(sock->sd, 1);
    }
 
    /* set socket to non-blocking mode */
    set_nonblock(sock->sd);
 
    /* set Path MTU discovery options on the socket */
    set_mtu_discover_type(sock->sd, sock->mtu_discover_type, sock->info.af);
 
#if EXTENDED_SOCKET_ERROR_CAPABILITY
    /* if the OS supports it, enable extended error passing on the socket */
    set_sock_extended_error_passing(sock->sd, sock->info.af);
#endif
}
 
 
static void
linksock_print_addr(struct link_socket *sock)
{
    struct gc_arena gc = gc_new();
    const msglvl_t msglevel = (sock->mode == LS_MODE_TCP_ACCEPT_FROM) ? D_INIT_MEDIUM : M_INFO;
 
    /* print local address */
    if (sock->bind_local)
    {
        sa_family_t ai_family = sock->info.lsa->actual.dest.addr.sa.sa_family;
        /* Socket is always bound on the first matching address,
         * For bound sockets with no remote addr this is the element of
         * the list */
        const struct addrinfo *cur;
        for (cur = sock->info.lsa->bind_local; cur; cur = cur->ai_next)
        {
            if (!ai_family || ai_family == cur->ai_family)
            {
                break;
            }
        }
        ASSERT(cur);
        msg(msglevel, "%s link local (bound): %s",
            proto2ascii(sock->info.proto, sock->info.af, true), print_sockaddr(cur->ai_addr, &gc));
    }
    else
    {
        msg(msglevel, "%s link local: (not bound)",
            proto2ascii(sock->info.proto, sock->info.af, true));
    }
 
    /* print active remote address */
    msg(msglevel, "%s link remote: %s", proto2ascii(sock->info.proto, sock->info.af, true),
        print_link_socket_actual_ex(&sock->info.lsa->actual, ":", PS_SHOW_PORT_IF_DEFINED, &gc));
    gc_free(&gc);
}
 
static void
phase2_tcp_server(struct link_socket *sock, struct signal_info *sig_info)
{
    ASSERT(sig_info);
    volatile int *signal_received = &sig_info->signal_received;
    switch (sock->mode)
    {
        case LS_MODE_DEFAULT:
            sock->sd =
                socket_listen_accept(sock->sd, &sock->info.lsa->actual,
                                     sock->info.lsa->bind_local, true, false,
                                     signal_received);
            break;
 
        case LS_MODE_TCP_LISTEN:
            socket_do_listen(sock->sd, sock->info.lsa->bind_local, true, false);
            break;
 
        case LS_MODE_TCP_ACCEPT_FROM:
            sock->sd = socket_do_accept(sock->sd, &sock->info.lsa->actual, false);
            if (!socket_defined(sock->sd))
            {
                register_signal(sig_info, SIGTERM, "socket-undefined");
                return;
            }
            tcp_connection_established(&sock->info.lsa->actual);
            break;
 
        default:
            ASSERT(0);
    }
}
 
 
static void
phase2_tcp_client(struct link_socket *sock, struct signal_info *sig_info)
{
    bool proxy_retry = false;
    do
    {
        socket_connect(&sock->sd, sock->info.lsa->current_remote->ai_addr,
                       get_server_poll_remaining_time(sock->server_poll_timeout), sig_info);
 
        if (sig_info->signal_received)
        {
            return;
        }
 
        if (sock->http_proxy)
        {
            proxy_retry = establish_http_proxy_passthru(
                sock->http_proxy, sock->sd, sock->proxy_dest_host, sock->proxy_dest_port,
                sock->server_poll_timeout, &sock->stream_buf.residual, sig_info);
        }
        else if (sock->socks_proxy)
        {
            establish_socks_proxy_passthru(sock->socks_proxy, sock->sd, sock->proxy_dest_host,
                                           sock->proxy_dest_port, sock->server_poll_timeout,
                                           sig_info);
        }
        if (proxy_retry)
        {
            openvpn_close_socket(sock->sd);
            sock->sd = create_socket_tcp(sock->info.lsa->current_remote);
        }
 
    } while (proxy_retry);
}
 
static void
phase2_socks_client(struct link_socket *sock, struct signal_info *sig_info)
{
    socket_connect(&sock->ctrl_sd, sock->info.lsa->current_remote->ai_addr,
                   get_server_poll_remaining_time(sock->server_poll_timeout), sig_info);
 
    if (sig_info->signal_received)
    {
        return;
    }
 
    establish_socks_proxy_udpassoc(sock->socks_proxy, sock->ctrl_sd, &sock->socks_relay.dest,
                                   sock->server_poll_timeout, sig_info);
 
    if (sig_info->signal_received)
    {
        return;
    }
 
    sock->remote_host = sock->proxy_dest_host;
    sock->remote_port = sock->proxy_dest_port;
 
    addr_zero_host(&sock->info.lsa->actual.dest);
    if (sock->info.lsa->remote_list)
    {
        freeaddrinfo(sock->info.lsa->remote_list);
        sock->info.lsa->current_remote = NULL;
        sock->info.lsa->remote_list = NULL;
    }
 
    resolve_remote(sock, 1, sig_info);
}
 
#if defined(_WIN32)
static void
create_socket_dco_win(struct context *c, struct link_socket *sock, struct signal_info *sig_info)
{
    /* in P2P mode we must have remote resolved at this point */
    const struct addrinfo *remoteaddr = sock->info.lsa->current_remote;
    if ((c->options.mode == MODE_POINT_TO_POINT) && (!remoteaddr))
    {
        return;
    }
 
    if (!c->c1.tuntap)
    {
        struct tuntap *tt;
        ALLOC_OBJ_CLEAR(tt, struct tuntap);
 
        tt->backend_driver = DRIVER_DCO;
        tt->options.msg_channel = c->options.msg_channel;
 
        const char *device_guid = NULL; /* not used */
        tun_open_device(tt, c->options.dev_node, &device_guid, &c->gc);
 
        /* Ensure we can "safely" cast the handle to a socket */
        static_assert(sizeof(sock->sd) == sizeof(tt->hand), "HANDLE and SOCKET size differs");
 
        c->c1.tuntap = tt;
    }
 
    if (c->options.mode == MODE_SERVER)
    {
        dco_mp_start_vpn(c->c1.tuntap->hand, sock);
    }
    else
    {
        dco_p2p_new_peer(c->c1.tuntap->hand, &c->c1.tuntap->dco_new_peer_ov, sock, sig_info);
    }
    sock->sockflags |= SF_DCO_WIN;
 
    if (sig_info->signal_received)
    {
        return;
    }
 
    sock->sd = (SOCKET)c->c1.tuntap->hand;
    linksock_print_addr(sock);
}
#endif /* if defined(_WIN32) */
 
/* finalize socket initialization */
void
link_socket_init_phase2(struct context *c, struct link_socket *sock)
{
    const struct frame *frame = &c->c2.frame;
    struct signal_info *sig_info = c->sig;
 
    struct signal_info sig_save = { 0 };
 
    ASSERT(sock);
    ASSERT(sig_info);
 
    if (sig_info->signal_received)
    {
        sig_save = *sig_info;
        sig_save.signal_received = signal_reset(sig_info, 0);
    }
 
    /* initialize buffers */
    socket_frame_init(frame, sock);
 
    /* Second chance to resolv/create socket */
    resolve_remote(sock, 2, sig_info);
 
    /* If a valid remote has been found, create the socket with its addrinfo */
#if defined(_WIN32)
    if (dco_enabled(&c->options))
    {
        create_socket_dco_win(c, sock, sig_info);
        goto done;
    }
#endif
    if (sock->info.lsa->current_remote)
    {
        create_socket(sock, sock->info.lsa->current_remote);
    }
 
    /* If socket has not already been created create it now */
    if (sock->sd == SOCKET_UNDEFINED)
    {
        /* If we have no --remote and have still not figured out the
         * protocol family to use we will use the first of the bind */
 
        if (sock->bind_local && !sock->remote_host && sock->info.lsa->bind_local)
        {
            /* Warn if this is because neither v4 or v6 was specified
             * and we should not connect a remote */
            if (sock->info.af == AF_UNSPEC)
            {
                sock->info.af = (sa_family_t)sock->info.lsa->bind_local->ai_family;
                msg(M_WARN, "Could not determine IPv4/IPv6 protocol. Using %s",
                    addr_family_name(sock->info.af));
            }
            create_socket(sock, sock->info.lsa->bind_local);
        }
    }
 
    /* Socket still undefined, give a warning and abort connection */
    if (sock->sd == SOCKET_UNDEFINED)
    {
        msg(M_WARN, "Could not determine IPv4/IPv6 protocol");
        register_signal(sig_info, SIGUSR1, "Could not determine IPv4/IPv6 protocol");
        goto done;
    }
 
    if (sig_info->signal_received)
    {
        goto done;
    }
 
    if (sock->info.proto == PROTO_TCP_SERVER)
    {
        phase2_tcp_server(sock, sig_info);
    }
    else if (sock->info.proto == PROTO_TCP_CLIENT)
    {
        phase2_tcp_client(sock, sig_info);
    }
    else if (sock->info.proto == PROTO_UDP && sock->socks_proxy)
    {
        phase2_socks_client(sock, sig_info);
    }
#ifdef TARGET_ANDROID
    if (sock->sd != -1)
    {
        protect_fd_nonlocal(sock->sd, &sock->info.lsa->actual.dest.addr.sa);
    }
#endif
    if (sig_info->signal_received)
    {
        goto done;
    }
 
    phase2_set_socket_flags(sock);
    linksock_print_addr(sock);
 
done:
    if (sig_save.signal_received)
    {
        /* Always restore the saved signal -- register/throw_signal will handle priority */
        if (sig_save.source == SIG_SOURCE_HARD && sig_info == &siginfo_static)
        {
            throw_signal(sig_save.signal_received);
        }
        else
        {
            register_signal(sig_info, sig_save.signal_received, sig_save.signal_text);
        }
    }
}
 
void
link_socket_close(struct link_socket *sock)
{
    if (sock)
    {
#ifdef ENABLE_DEBUG
        const int gremlin = GREMLIN_CONNECTION_FLOOD_LEVEL(sock->gremlin);
#else
        const int gremlin = 0;
#endif
 
        if (socket_defined(sock->sd))
        {
#ifdef _WIN32
            close_net_event_win32(&sock->listen_handle, sock->sd, 0);
#endif
            if (!gremlin)
            {
                msg(D_LOW, "TCP/UDP: Closing socket");
                if (openvpn_close_socket(sock->sd))
                {
                    msg(M_WARN | M_ERRNO, "TCP/UDP: Close Socket failed");
                }
            }
            sock->sd = SOCKET_UNDEFINED;
#ifdef _WIN32
            if (!gremlin)
            {
                overlapped_io_close(&sock->reads);
                overlapped_io_close(&sock->writes);
            }
#endif
        }
 
        if (socket_defined(sock->ctrl_sd))
        {
            if (openvpn_close_socket(sock->ctrl_sd))
            {
                msg(M_WARN | M_ERRNO, "TCP/UDP: Close Socket (ctrl_sd) failed");
            }
            sock->ctrl_sd = SOCKET_UNDEFINED;
        }
 
        stream_buf_close(&sock->stream_buf);
        free_buf(&sock->stream_buf_data);
        if (!gremlin)
        {
            free(sock);
        }
    }
}
 
void
setenv_trusted(struct env_set *es, const struct link_socket_info *info)
{
    setenv_link_socket_actual(es, "trusted", &info->lsa->actual, SA_IP_PORT);
}
 
static void
ipchange_fmt(const bool include_cmd, struct argv *argv, const struct link_socket_info *info,
             struct gc_arena *gc)
{
    const char *host = print_sockaddr_ex(&info->lsa->actual.dest.addr.sa, " ", PS_SHOW_PORT, gc);
    if (include_cmd)
    {
        argv_parse_cmd(argv, info->ipchange_command);
        argv_printf_cat(argv, "%s", host);
    }
    else
    {
        argv_printf(argv, "%s", host);
    }
}
 
void
link_socket_connection_initiated(struct link_socket_info *info,
                                 const struct link_socket_actual *act, const char *common_name,
                                 struct env_set *es)
{
    struct gc_arena gc = gc_new();
 
    info->lsa->actual = *act; /* Note: skip this line for --force-dest */
    setenv_trusted(es, info);
    info->connection_established = true;
 
    /* Print connection initiated message, with common name if available */
    {
        struct buffer out = alloc_buf_gc(256, &gc);
        if (common_name)
        {
            buf_printf(&out, "[%s] ", common_name);
        }
        buf_printf(&out, "Peer Connection Initiated with %s",
                   print_link_socket_actual(&info->lsa->actual, &gc));
        msg(M_INFO, "%s", BSTR(&out));
    }
 
    /* set environmental vars */
    setenv_str(es, "common_name", common_name);
 
    /* Process --ipchange plugin */
    if (plugin_defined(info->plugins, OPENVPN_PLUGIN_IPCHANGE))
    {
        struct argv argv = argv_new();
        ipchange_fmt(false, &argv, info, &gc);
        if (plugin_call(info->plugins, OPENVPN_PLUGIN_IPCHANGE, &argv, NULL, es)
            != OPENVPN_PLUGIN_FUNC_SUCCESS)
        {
            msg(M_WARN, "WARNING: ipchange plugin call failed");
        }
        argv_free(&argv);
    }
 
    /* Process --ipchange option */
    if (info->ipchange_command)
    {
        struct argv argv = argv_new();
        setenv_str(es, "script_type", "ipchange");
        ipchange_fmt(true, &argv, info, &gc);
        openvpn_run_script(&argv, es, 0, "--ipchange");
        argv_free(&argv);
    }
 
    gc_free(&gc);
}
 
void
link_socket_bad_incoming_addr(struct buffer *buf, const struct link_socket_info *info,
                              const struct link_socket_actual *from_addr)
{
    struct gc_arena gc = gc_new();
    const struct addrinfo *ai;
 
    switch (from_addr->dest.addr.sa.sa_family)
    {
        case AF_INET:
        case AF_INET6:
            msg(D_LINK_ERRORS,
                "TCP/UDP: Incoming packet rejected from %s[%d], expected peer address: %s (allow this incoming source address/port by removing --remote or adding --float)",
                print_link_socket_actual(from_addr, &gc), (int)from_addr->dest.addr.sa.sa_family,
                print_sockaddr_ex(info->lsa->remote_list->ai_addr, ":", PS_SHOW_PORT, &gc));
            /* print additional remote addresses */
            for (ai = info->lsa->remote_list->ai_next; ai; ai = ai->ai_next)
            {
                msg(D_LINK_ERRORS, "or from peer address: %s",
                    print_sockaddr_ex(ai->ai_addr, ":", PS_SHOW_PORT, &gc));
            }
            break;
    }
    buf->len = 0;
    gc_free(&gc);
}
 
void
link_socket_bad_outgoing_addr(void)
{
    dmsg(D_READ_WRITE, "TCP/UDP: No outgoing address to send packet");
}
 
in_addr_t
link_socket_current_remote(const struct link_socket_info *info)
{
    const struct link_socket_addr *lsa = info->lsa;
 
    /*
     * This logic supports "redirect-gateway" semantic, which
     * makes sense only for PF_INET routes over PF_INET endpoints
     *
     * Maybe in the future consider PF_INET6 endpoints also ...
     * by now just ignore it
     *
     * For --remote entries with multiple addresses this
     * only return the actual endpoint we have successfully connected to
     */
    if (lsa->actual.dest.addr.sa.sa_family != AF_INET)
    {
        return IPV4_INVALID_ADDR;
    }
 
    if (link_socket_actual_defined(&lsa->actual))
    {
        return ntohl(lsa->actual.dest.addr.in4.sin_addr.s_addr);
    }
    else if (lsa->current_remote)
    {
        return ntohl(((struct sockaddr_in *)lsa->current_remote->ai_addr)->sin_addr.s_addr);
    }
    else
    {
        return 0;
    }
}
 
const struct in6_addr *
link_socket_current_remote_ipv6(const struct link_socket_info *info)
{
    const struct link_socket_addr *lsa = info->lsa;
 
    /* This logic supports "redirect-gateway" semantic,
     * for PF_INET6 routes over PF_INET6 endpoints
     *
     * For --remote entries with multiple addresses this
     * only return the actual endpoint we have successfully connected to
     */
    if (lsa->actual.dest.addr.sa.sa_family != AF_INET6)
    {
        return NULL;
    }
 
    if (link_socket_actual_defined(&lsa->actual))
    {
        return &(lsa->actual.dest.addr.in6.sin6_addr);
    }
    else if (lsa->current_remote)
    {
        return &(((struct sockaddr_in6 *)lsa->current_remote->ai_addr)->sin6_addr);
    }
    else
    {
        return NULL;
    }
}
 
/*
 * Return a status string describing socket state.
 */
const char *
socket_stat(const struct link_socket *s, unsigned int rwflags, struct gc_arena *gc)
{
    struct buffer out = alloc_buf_gc(64, gc);
    if (s)
    {
        if (rwflags & EVENT_READ)
        {
            buf_printf(&out, "S%s", (s->rwflags_debug & EVENT_READ) ? "R" : "r");
#ifdef _WIN32
            buf_printf(&out, "%s", overlapped_io_state_ascii(&s->reads));
#endif
        }
        if (rwflags & EVENT_WRITE)
        {
            buf_printf(&out, "S%s", (s->rwflags_debug & EVENT_WRITE) ? "W" : "w");
#ifdef _WIN32
            buf_printf(&out, "%s", overlapped_io_state_ascii(&s->writes));
#endif
        }
    }
    else
    {
        buf_printf(&out, "S?");
    }
    return BSTR(&out);
}
 
/*
 * Stream buffer functions, used to packetize a TCP
 * stream connection.
 */

static inline void
stream_buf_reset(struct stream_buf *sb)
{
    dmsg(D_STREAM_DEBUG, "STREAM: RESET");
    sb->residual_fully_formed = false;
    sb->buf = sb->buf_init;
    sb->len = -1;
}
 
static void
stream_buf_init(struct stream_buf *sb, const struct buffer *buf, const unsigned int sockflags,
                const int proto)
{
    sb->buf_init = *buf;
    sb->maxlen = sb->buf_init.len;
    sb->buf_init.len = 0;
    sb->residual = alloc_buf(sb->maxlen);
    sb->error = false;
#if PORT_SHARE
    sb->port_share_state =
        ((sockflags & SF_PORT_SHARE) && (proto == PROTO_TCP_SERVER)) ? PS_ENABLED : PS_DISABLED;
#endif
    stream_buf_reset(sb);
 
    dmsg(D_STREAM_DEBUG, "STREAM: INIT maxlen=%d", sb->maxlen);
}

static inline struct buffer
stream_buf_get_next(struct stream_buf *sb)
{
    /* set up 'next' for next i/o read */
    struct buffer next;
    next = sb->buf;
    next.offset = sb->buf.offset + sb->buf.len;
    next.len = (sb->len >= 0 ? sb->len : sb->maxlen) - sb->buf.len;
    dmsg(D_STREAM_DEBUG, "STREAM: GET NEXT, buf=[%d,%d] next=[%d,%d] len=%d maxlen=%d",
         sb->buf.offset, sb->buf.len, next.offset, next.len, sb->len, sb->maxlen);
    ASSERT(next.len > 0);
    ASSERT(buf_safe(&sb->buf, next.len));
    return next;
}

static inline void
stream_buf_get_final(struct stream_buf *sb, struct buffer *buf)
{
    dmsg(D_STREAM_DEBUG, "STREAM: GET FINAL len=%d", buf_defined(&sb->buf) ? sb->buf.len : -1);
    ASSERT(buf_defined(&sb->buf));
    *buf = sb->buf;
}
 
bool
stream_buf_read_setup_dowork(struct stream_buf *sb)
{
    if (sb->residual.len && !sb->residual_fully_formed)
    {
        ASSERT(buf_copy(&sb->buf, &sb->residual));
        ASSERT(buf_init(&sb->residual, 0));
        sb->residual_fully_formed = stream_buf_added(sb, 0);
        dmsg(D_STREAM_DEBUG, "STREAM: RESIDUAL FULLY FORMED [%s], len=%d",
             sb->residual_fully_formed ? "YES" : "NO", sb->residual.len);
    }
 
    return !sb->residual_fully_formed;
}

static bool
stream_buf_added(struct stream_buf *sb, ssize_t length_added)
{
    dmsg(D_STREAM_DEBUG, "STREAM: ADD length_added=%zd", length_added);
    if (length_added > 0)
    {
        ASSERT(sb->buf.len + length_added <= INT_MAX);
        sb->buf.len += (int)length_added;
    }
 
    /* if length unknown, see if we can get the length prefix from
     * the head of the buffer */
    if (sb->len < 0 && sb->buf.len >= (int)sizeof(packet_size_type))
    {
        packet_size_type net_size;
 
#if PORT_SHARE
        if (sb->port_share_state == PS_ENABLED)
        {
            if (!is_openvpn_protocol(&sb->buf))
            {
                msg(D_PS_PROXY, "Non-OpenVPN client protocol detected");
                sb->port_share_state = PS_FOREIGN;
                sb->error = true;
                return false;
            }
            else
            {
                sb->port_share_state = PS_DISABLED;
            }
        }
#endif
 
        ASSERT(buf_read(&sb->buf, &net_size, sizeof(net_size)));
        sb->len = ntohps(net_size);
 
        if (sb->len < 1 || sb->len > sb->maxlen)
        {
            msg(M_WARN,
                "WARNING: Bad encapsulated packet length from peer (%d), which must be > 0 and <= %d -- please ensure that --tun-mtu or --link-mtu is equal on both peers -- this condition could also indicate a possible active attack on the TCP link -- [Attempting restart...]",
                sb->len, sb->maxlen);
            stream_buf_reset(sb);
            sb->error = true;
            return false;
        }
    }
 
    /* is our incoming packet fully read? */
    if (sb->len > 0 && sb->buf.len >= sb->len)
    {
        /* save any residual data that's part of the next packet */
        ASSERT(buf_init(&sb->residual, 0));
        if (sb->buf.len > sb->len)
        {
            ASSERT(buf_copy_excess(&sb->residual, &sb->buf, sb->len));
        }
        dmsg(D_STREAM_DEBUG, "STREAM: ADD returned TRUE, buf_len=%d, residual_len=%d",
             BLEN(&sb->buf), BLEN(&sb->residual));
        return true;
    }
    else
    {
        dmsg(D_STREAM_DEBUG, "STREAM: ADD returned FALSE (have=%d need=%d)", sb->buf.len, sb->len);
        return false;
    }
}
 
static void
stream_buf_close(struct stream_buf *sb)
{
    free_buf(&sb->residual);
}
 
/*
 * The listen event is a special event whose sole purpose is
 * to tell us that there's a new incoming connection on a
 * TCP socket, for use in server mode.
 */
event_t
socket_listen_event_handle(struct link_socket *s)
{
#ifdef _WIN32
    if (!defined_net_event_win32(&s->listen_handle))
    {
        init_net_event_win32(&s->listen_handle, FD_ACCEPT, s->sd, 0);
    }
    return &s->listen_handle;
#else /* ifdef _WIN32 */
    return s->sd;
#endif
}
 
 
/*
 * Bad incoming address lengths that differ from what
 * we expect are considered to be fatal errors.
 */
void
bad_address_length(int actual, int expected)
{
    msg(M_FATAL,
        "ERROR: received strange incoming packet with an address length of %d -- we only accept address lengths of %d.",
        actual, expected);
}
 
/*
 * Socket Read Routines
 */
 
ssize_t
link_socket_read_tcp(struct link_socket *sock, struct buffer *buf)
{
    ssize_t len = 0;
 
    if (!sock->stream_buf.residual_fully_formed)
    {
        /* with Linux-DCO, we sometimes try to access a socket that is
         * already installed in the kernel and has no valid file descriptor
         * anymore.  This is a bug.
         * Handle by resetting client instance instead of crashing.
         */
        if (sock->sd == SOCKET_UNDEFINED)
        {
            msg(M_INFO, "BUG: link_socket_read_tcp(): sock->sd==-1, reset client instance");
            sock->stream_reset = true; /* reset client instance */
            return buf->len = 0;       /* nothing to read */
        }
 
#ifdef _WIN32
        sockethandle_t sh = { .s = sock->sd };
        len = sockethandle_finalize(sh, &sock->reads, buf, NULL);
#else
        struct buffer frag = stream_buf_get_next(&sock->stream_buf);
        len = recv(sock->sd, BPTR(&frag), BLENZ(&frag), MSG_NOSIGNAL);
#endif
 
        if (!len)
        {
            sock->stream_reset = true;
        }
        if (len <= 0)
        {
            buf->len = 0;
            return len;
        }
    }
 
    if (sock->stream_buf.residual_fully_formed
        || stream_buf_added(&sock->stream_buf, len)) /* packet complete? */
    {
        stream_buf_get_final(&sock->stream_buf, buf);
        stream_buf_reset(&sock->stream_buf);
        return buf->len;
    }
    else
    {
        return buf->len = 0; /* no error, but packet is still incomplete */
    }
}
 
#ifndef _WIN32
 
#if ENABLE_IP_PKTINFO
 
/* make the buffer large enough to handle ancillary socket data for
 * both IPv4 and IPv6 destination addresses, plus padding (see RFC 2292)
 */
#if defined(HAVE_IN_PKTINFO) && defined(HAVE_IPI_SPEC_DST)
#define PKTINFO_BUF_SIZE \
    max_int(CMSG_SPACE(sizeof(struct in6_pktinfo)), CMSG_SPACE(sizeof(struct in_pktinfo)))
#else
#define PKTINFO_BUF_SIZE \
    max_int(CMSG_SPACE(sizeof(struct in6_pktinfo)), CMSG_SPACE(sizeof(struct in_addr)))
#endif
 
static ssize_t
link_socket_read_udp_posix_recvmsg(const struct link_socket *sock, struct buffer *buf,
                                   struct link_socket_actual *from, socklen_t *fromlen)
{
    struct iovec iov;
    uint8_t pktinfo_buf[PKTINFO_BUF_SIZE];
    struct msghdr mesg = { 0 };
    *fromlen = sizeof(from->dest.addr);
 
    ASSERT(sock->sd >= 0); /* can't happen */
 
    iov.iov_base = BPTR(buf);
    iov.iov_len = buf_forward_capacity_total(buf);
    mesg.msg_iov = &iov;
    mesg.msg_iovlen = 1;
    mesg.msg_name = &from->dest.addr;
    mesg.msg_namelen = *fromlen;
    mesg.msg_control = pktinfo_buf;
    mesg.msg_controllen = (socklen_t)sizeof(pktinfo_buf);
    ssize_t len = recvmsg(sock->sd, &mesg, 0);
    if (len < 0)
    {
        buf->len = 0;
        return len;
    }
    ASSERT(len <= INT_MAX);
    buf->len = (int)len;
    struct cmsghdr *cmsg;
    *fromlen = mesg.msg_namelen;
    cmsg = CMSG_FIRSTHDR(&mesg);
    if (cmsg != NULL && CMSG_NXTHDR(&mesg, cmsg) == NULL
#if defined(HAVE_IN_PKTINFO) && defined(HAVE_IPI_SPEC_DST)
        && cmsg->cmsg_level == SOL_IP && cmsg->cmsg_type == IP_PKTINFO
        && cmsg->cmsg_len >= CMSG_LEN(sizeof(struct in_pktinfo)))
#elif defined(IP_RECVDSTADDR)
        && cmsg->cmsg_level == IPPROTO_IP && cmsg->cmsg_type == IP_RECVDSTADDR
        && cmsg->cmsg_len >= CMSG_LEN(sizeof(struct in_addr)))
#else /* if defined(HAVE_IN_PKTINFO) && defined(HAVE_IPI_SPEC_DST) */
#error ENABLE_IP_PKTINFO is set without IP_PKTINFO xor IP_RECVDSTADDR (fix syshead.h)
#endif
    {
#if defined(HAVE_IN_PKTINFO) && defined(HAVE_IPI_SPEC_DST)
        struct in_pktinfo *pkti = (struct in_pktinfo *)CMSG_DATA(cmsg);
        from->pi.in4.ipi_ifindex =
            (sock->sockflags & SF_PKTINFO_COPY_IIF) ? pkti->ipi_ifindex : 0;
        from->pi.in4.ipi_spec_dst = pkti->ipi_spec_dst;
#elif defined(IP_RECVDSTADDR)
        from->pi.in4 = *(struct in_addr *)CMSG_DATA(cmsg);
#else /* if defined(HAVE_IN_PKTINFO) && defined(HAVE_IPI_SPEC_DST) */
#error ENABLE_IP_PKTINFO is set without IP_PKTINFO xor IP_RECVDSTADDR (fix syshead.h)
#endif
    }
    else if (cmsg != NULL && CMSG_NXTHDR(&mesg, cmsg) == NULL
             && cmsg->cmsg_level == IPPROTO_IPV6 && cmsg->cmsg_type == IPV6_PKTINFO
             && cmsg->cmsg_len >= CMSG_LEN(sizeof(struct in6_pktinfo)))
    {
        struct in6_pktinfo *pkti6 = (struct in6_pktinfo *)CMSG_DATA(cmsg);
        from->pi.in6.ipi6_ifindex =
            (sock->sockflags & SF_PKTINFO_COPY_IIF) ? pkti6->ipi6_ifindex : 0;
        from->pi.in6.ipi6_addr = pkti6->ipi6_addr;
    }
    else if (cmsg != NULL)
    {
        msg(M_WARN,
            "CMSG received that cannot be parsed (cmsg_level=%d, cmsg_type=%d, cmsg=len=%zu)",
            cmsg->cmsg_level, cmsg->cmsg_type, (size_t)cmsg->cmsg_len);
    }
 
    return buf->len;
}
#endif /* if ENABLE_IP_PKTINFO */
 
ssize_t
link_socket_read_udp_posix(const struct link_socket *sock, struct buffer *buf,
                           struct link_socket_actual *from)
{
    ssize_t recvlen;
    socklen_t fromlen = sizeof(from->dest.addr);
    socklen_t expectedlen = af_addr_size(sock->info.af);
    addr_zero_host(&from->dest);
 
    ASSERT(sock->sd >= 0); /* can't happen */
 
#if ENABLE_IP_PKTINFO
    /* Both PROTO_UDPv4 and PROTO_UDPv6 */
    if (sock->info.proto == PROTO_UDP && sock->sockflags & SF_USE_IP_PKTINFO)
    {
        recvlen = link_socket_read_udp_posix_recvmsg(sock, buf, from, &fromlen);
    }
    else
#endif
    {
        recvlen = recvfrom(sock->sd, BPTR(buf), buf_forward_capacity(buf), 0,
                           &from->dest.addr.sa, &fromlen);
    }
    if (recvlen < 0)
    {
        buf->len = 0;
        return recvlen;
    }
    ASSERT(recvlen <= INT_MAX);
    buf->len = (int)recvlen;
    /* FIXME: won't do anything when sock->info.af == AF_UNSPEC */
    if (expectedlen && fromlen != expectedlen)
    {
        bad_address_length(fromlen, expectedlen);
    }
    return buf->len;
}
 
#if defined(__GNUC__) || defined(__clang__)
#pragma GCC diagnostic pop
#endif
 
#endif /* ifndef _WIN32 */
 
/*
 * Socket Write Routines
 */
 
ssize_t
link_socket_write_tcp(struct link_socket *sock, struct buffer *buf, struct link_socket_actual *to)
{
    const int blen = BLEN(buf);
    ASSERT(blen >= 0 && blen <= PACKET_SIZE_MAX);
    packet_size_type len = (packet_size_type)blen;
    dmsg(D_STREAM_DEBUG, "STREAM: WRITE %u offset=%d", len, buf->offset);
    ASSERT(len <= sock->stream_buf.maxlen);
    len = htonps(len);
    ASSERT(buf_write_prepend(buf, &len, sizeof(len)));
#ifdef _WIN32
    return link_socket_write_win32(sock, buf, to);
#else
    return link_socket_write_tcp_posix(sock, buf);
#endif
}
 
#if ENABLE_IP_PKTINFO
 
ssize_t
link_socket_write_udp_posix_sendmsg(const struct link_socket *sock, const struct buffer *buf,
                                    struct link_socket_actual *to)
{
    struct iovec iov;
    struct msghdr mesg;
    struct cmsghdr *cmsg;
    uint8_t pktinfo_buf[PKTINFO_BUF_SIZE];
 
    /* sendmsg takes a const msghdr, but we can't construct that here
       directly, so cast */
    iov.iov_base = (char *)CBPTR(buf);
    iov.iov_len = BLENZ(buf);
    mesg.msg_iov = &iov;
    mesg.msg_iovlen = 1;
    switch (to->dest.addr.sa.sa_family)
    {
        case AF_INET:
        {
            mesg.msg_name = &to->dest.addr.sa;
            mesg.msg_namelen = sizeof(struct sockaddr_in);
            mesg.msg_control = pktinfo_buf;
            mesg.msg_flags = 0;
#if defined(HAVE_IN_PKTINFO) && defined(HAVE_IPI_SPEC_DST)
            mesg.msg_controllen = CMSG_SPACE(sizeof(struct in_pktinfo));
            cmsg = CMSG_FIRSTHDR(&mesg);
            cmsg->cmsg_len = CMSG_LEN(sizeof(struct in_pktinfo));
            cmsg->cmsg_level = SOL_IP;
            cmsg->cmsg_type = IP_PKTINFO;
            {
                struct in_pktinfo *pkti;
                pkti = (struct in_pktinfo *)CMSG_DATA(cmsg);
                pkti->ipi_ifindex = to->pi.in4.ipi_ifindex;
                pkti->ipi_spec_dst = to->pi.in4.ipi_spec_dst;
                pkti->ipi_addr.s_addr = 0;
            }
#elif defined(IP_RECVDSTADDR)
            ASSERT(CMSG_SPACE(sizeof(struct in_addr)) <= sizeof(pktinfo_buf));
            mesg.msg_controllen = CMSG_SPACE(sizeof(struct in_addr));
            cmsg = CMSG_FIRSTHDR(&mesg);
            cmsg->cmsg_len = CMSG_LEN(sizeof(struct in_addr));
            cmsg->cmsg_level = IPPROTO_IP;
            cmsg->cmsg_type = IP_RECVDSTADDR;
            *(struct in_addr *)CMSG_DATA(cmsg) = to->pi.in4;
#else  /* if defined(HAVE_IN_PKTINFO) && defined(HAVE_IPI_SPEC_DST) */
#error ENABLE_IP_PKTINFO is set without IP_PKTINFO xor IP_RECVDSTADDR (fix syshead.h)
#endif /* if defined(HAVE_IN_PKTINFO) && defined(HAVE_IPI_SPEC_DST) */
            break;
        }
 
        case AF_INET6:
        {
            struct in6_pktinfo *pkti6;
            mesg.msg_name = &to->dest.addr.sa;
            mesg.msg_namelen = sizeof(struct sockaddr_in6);
 
            ASSERT(CMSG_SPACE(sizeof(struct in6_pktinfo)) <= sizeof(pktinfo_buf));
            mesg.msg_control = pktinfo_buf;
            mesg.msg_controllen = CMSG_SPACE(sizeof(struct in6_pktinfo));
            mesg.msg_flags = 0;
            cmsg = CMSG_FIRSTHDR(&mesg);
            cmsg->cmsg_len = CMSG_LEN(sizeof(struct in6_pktinfo));
            cmsg->cmsg_level = IPPROTO_IPV6;
            cmsg->cmsg_type = IPV6_PKTINFO;
 
            pkti6 = (struct in6_pktinfo *)CMSG_DATA(cmsg);
            pkti6->ipi6_ifindex = to->pi.in6.ipi6_ifindex;
            pkti6->ipi6_addr = to->pi.in6.ipi6_addr;
            break;
        }
 
        default:
            ASSERT(0);
    }
    return sendmsg(sock->sd, &mesg, 0);
}
 
#endif /* if ENABLE_IP_PKTINFO */
 
/*
 * Win32 overlapped socket I/O functions.
 */
 
#ifdef _WIN32
 
static int
socket_get_last_error(const struct link_socket *sock)
{
    if (socket_is_dco_win(sock))
    {
        return GetLastError();
    }
 
    return WSAGetLastError();
}
 
int
socket_recv_queue(struct link_socket *sock, int maxsize)
{
    if (sock->reads.iostate == IOSTATE_INITIAL)
    {
        WSABUF wsabuf[1];
        int status;
 
        /* reset buf to its initial state */
        if (proto_is_udp(sock->info.proto))
        {
            sock->reads.buf = sock->reads.buf_init;
        }
        else if (proto_is_tcp(sock->info.proto))
        {
            sock->reads.buf = stream_buf_get_next(&sock->stream_buf);
        }
        else
        {
            ASSERT(0);
        }
 
        /* Win32 docs say it's okay to allocate the wsabuf on the stack */
        wsabuf[0].buf = BSTR(&sock->reads.buf);
        /* make sure maxsize is sane */
        ASSERT(maxsize <= BLEN(&sock->reads.buf));
        wsabuf[0].len = maxsize ? maxsize : BLEN(&sock->reads.buf);
 
        /* the overlapped read will signal this event on I/O completion */
        ASSERT(ResetEvent(sock->reads.overlapped.hEvent));
        sock->reads.flags = 0;
 
        if (socket_is_dco_win(sock))
        {
            status = ReadFile((HANDLE)sock->sd, wsabuf[0].buf, wsabuf[0].len, &sock->reads.size,
                              &sock->reads.overlapped);
            /* Readfile status is inverted from WSARecv */
            status = !status;
        }
        else if (proto_is_udp(sock->info.proto))
        {
            sock->reads.addr_defined = true;
            sock->reads.addrlen = sizeof(sock->reads.addr6);
            status = WSARecvFrom(sock->sd, wsabuf, 1, &sock->reads.size, &sock->reads.flags,
                                 (struct sockaddr *)&sock->reads.addr, &sock->reads.addrlen,
                                 &sock->reads.overlapped, NULL);
        }
        else if (proto_is_tcp(sock->info.proto))
        {
            sock->reads.addr_defined = false;
            status = WSARecv(sock->sd, wsabuf, 1, &sock->reads.size, &sock->reads.flags,
                             &sock->reads.overlapped, NULL);
        }
        else
        {
            status = 0;
            ASSERT(0);
        }
 
        if (!status) /* operation completed immediately? */
        {
            /* FIXME: won't do anything when sock->info.af == AF_UNSPEC */
            int af_len = af_addr_size(sock->info.af);
            if (sock->reads.addr_defined && af_len && sock->reads.addrlen != af_len)
            {
                bad_address_length(sock->reads.addrlen, af_len);
            }
            sock->reads.iostate = IOSTATE_IMMEDIATE_RETURN;
 
            /* since we got an immediate return, we must signal the event object ourselves */
            ASSERT(SetEvent(sock->reads.overlapped.hEvent));
            sock->reads.status = 0;
 
            dmsg(D_WIN32_IO, "WIN32 I/O: Socket Receive immediate return [%d,%d]",
                 (int)wsabuf[0].len, (int)sock->reads.size);
        }
        else
        {
            status = socket_get_last_error(sock);
            if (status == WSA_IO_PENDING) /* operation queued? */
            {
                sock->reads.iostate = IOSTATE_QUEUED;
                sock->reads.status = status;
                dmsg(D_WIN32_IO, "WIN32 I/O: Socket Receive queued [%d]", (int)wsabuf[0].len);
            }
            else /* error occurred */
            {
                struct gc_arena gc = gc_new();
                ASSERT(SetEvent(sock->reads.overlapped.hEvent));
                sock->reads.iostate = IOSTATE_IMMEDIATE_RETURN;
                sock->reads.status = status;
                dmsg(D_WIN32_IO, "WIN32 I/O: Socket Receive error [%d]: %s", (int)wsabuf[0].len,
                     strerror_win32(status, &gc));
                gc_free(&gc);
            }
        }
    }
    return sock->reads.iostate;
}
 
int
socket_send_queue(struct link_socket *sock, const struct buffer *buf, const struct link_socket_actual *to)
{
    if (sock->writes.iostate == IOSTATE_INITIAL)
    {
        WSABUF wsabuf[1];
        int status;
 
        /* make a private copy of buf */
        sock->writes.buf = sock->writes.buf_init;
        sock->writes.buf.len = 0;
        ASSERT(buf_copy(&sock->writes.buf, buf));
 
        /* Win32 docs say it's okay to allocate the wsabuf on the stack */
        wsabuf[0].buf = BSTR(&sock->writes.buf);
        wsabuf[0].len = BLEN(&sock->writes.buf);
 
        /* the overlapped write will signal this event on I/O completion */
        ASSERT(ResetEvent(sock->writes.overlapped.hEvent));
        sock->writes.flags = 0;
 
        if (socket_is_dco_win(sock))
        {
            status = WriteFile((HANDLE)sock->sd, wsabuf[0].buf, wsabuf[0].len, &sock->writes.size,
                               &sock->writes.overlapped);
 
            /* WriteFile status is inverted from WSASendTo */
            status = !status;
        }
        else if (proto_is_udp(sock->info.proto))
        {
            /* set destination address for UDP writes */
            sock->writes.addr_defined = true;
            if (to->dest.addr.sa.sa_family == AF_INET6)
            {
                sock->writes.addr6 = to->dest.addr.in6;
                sock->writes.addrlen = sizeof(sock->writes.addr6);
            }
            else
            {
                sock->writes.addr = to->dest.addr.in4;
                sock->writes.addrlen = sizeof(sock->writes.addr);
            }
 
            status = WSASendTo(sock->sd, wsabuf, 1, &sock->writes.size, sock->writes.flags,
                               (struct sockaddr *)&sock->writes.addr, sock->writes.addrlen,
                               &sock->writes.overlapped, NULL);
        }
        else if (proto_is_tcp(sock->info.proto))
        {
            /* destination address for TCP writes was established on connection initiation */
            sock->writes.addr_defined = false;
 
            status = WSASend(sock->sd, wsabuf, 1, &sock->writes.size, sock->writes.flags,
                             &sock->writes.overlapped, NULL);
        }
        else
        {
            status = 0;
            ASSERT(0);
        }
 
        if (!status) /* operation completed immediately? */
        {
            sock->writes.iostate = IOSTATE_IMMEDIATE_RETURN;
 
            /* since we got an immediate return, we must signal the event object ourselves */
            ASSERT(SetEvent(sock->writes.overlapped.hEvent));
 
            sock->writes.status = 0;
 
            dmsg(D_WIN32_IO, "WIN32 I/O: Socket Send immediate return [%d,%d]", (int)wsabuf[0].len,
                 (int)sock->writes.size);
        }
        else
        {
            status = socket_get_last_error(sock);
            /* both status code have the identical value */
            if (status == WSA_IO_PENDING || status == ERROR_IO_PENDING) /* operation queued? */
            {
                sock->writes.iostate = IOSTATE_QUEUED;
                sock->writes.status = status;
                dmsg(D_WIN32_IO, "WIN32 I/O: Socket Send queued [%d]", (int)wsabuf[0].len);
            }
            else /* error occurred */
            {
                struct gc_arena gc = gc_new();
                ASSERT(SetEvent(sock->writes.overlapped.hEvent));
                sock->writes.iostate = IOSTATE_IMMEDIATE_RETURN;
                sock->writes.status = status;
 
                dmsg(D_WIN32_IO, "WIN32 I/O: Socket Send error [%d]: %s", (int)wsabuf[0].len,
                     strerror_win32(status, &gc));
 
                gc_free(&gc);
            }
        }
    }
    return sock->writes.iostate;
}
 
static void
read_sockaddr_from_overlapped(const struct overlapped_io *io, struct sockaddr *dst, int overlapped_ret)
{
    if (overlapped_ret >= 0 && io->addr_defined)
    {
        /* TODO(jjo): streamline this mess */
        /* in this func we don't have relevant info about the PF_ of this
         * endpoint, as link_socket_actual will be zero for the 1st received packet
         *
         * Test for inets PF_ possible sizes
         */
        switch (io->addrlen)
        {
            case sizeof(struct sockaddr_in):
            case sizeof(struct sockaddr_in6):
            /* TODO(jjo): for some reason (?) I'm getting 24,28 for AF_INET6
             * under _WIN32*/
            case sizeof(struct sockaddr_in6) - 4:
                break;
 
            default:
                bad_address_length(io->addrlen, af_addr_size(io->addr.sin_family));
        }
 
        switch (io->addr.sin_family)
        {
            case AF_INET:
                memcpy(dst, &io->addr, sizeof(struct sockaddr_in));
                break;
 
            case AF_INET6:
                memcpy(dst, &io->addr6, sizeof(struct sockaddr_in6));
                break;
        }
    }
    else
    {
        CLEAR(*dst);
    }
}

static int
read_sockaddr_from_packet(struct buffer *buf, struct sockaddr *dst)
{
    int sa_len = 0;
 
    /* In dco-win multipeer mode the kernel driver always prepends a full
     * sockaddr_in or sockaddr_in6 in front of the control-packet payload,
     * so the buffer must hold at least sizeof(struct sockaddr_in) bytes
     * before we may inspect sa_family. */
    ASSERT(buf_len(buf) >= (int)sizeof(struct sockaddr_in));
 
    const struct sockaddr *sa = (const struct sockaddr *)BPTR(buf);
    switch (sa->sa_family)
    {
        case AF_INET:
            sa_len = sizeof(struct sockaddr_in);
            break;
 
        case AF_INET6:
            sa_len = sizeof(struct sockaddr_in6);
            ASSERT(buf_len(buf) >= sa_len);
            break;
 
        default:
            ASSERT(0); /* driver validates the family before writing */
    }
 
    memcpy(dst, sa, sa_len);
    buf_advance(buf, sa_len);
 
    return sa_len;
}
 
/* Returns the number of bytes successfully read */
int
sockethandle_finalize(sockethandle_t sh, struct overlapped_io *io, struct buffer *buf,
                      struct link_socket_actual *from)
{
    int ret = -1;
    BOOL status;
 
    switch (io->iostate)
    {
        case IOSTATE_QUEUED:
            status = SocketHandleGetOverlappedResult(sh, io);
            if (status)
            {
                /* successful return for a queued operation */
                if (buf)
                {
                    *buf = io->buf;
                }
                ret = io->size;
                io->iostate = IOSTATE_INITIAL;
                ASSERT(ResetEvent(io->overlapped.hEvent));
 
                dmsg(D_WIN32_IO, "WIN32 I/O: Completion success [%d]", ret);
            }
            else
            {
                /* error during a queued operation */
                ret = -1;
                if (SocketHandleGetLastError(sh) != ERROR_IO_INCOMPLETE)
                {
                    /* if no error (i.e. just not finished yet), then DON'T execute this code */
                    io->iostate = IOSTATE_INITIAL;
                    ASSERT(ResetEvent(io->overlapped.hEvent));
                    msg(D_WIN32_IO | M_ERRNO, "WIN32 I/O: Completion error");
                }
            }
            break;
 
        case IOSTATE_IMMEDIATE_RETURN:
            io->iostate = IOSTATE_INITIAL;
            ASSERT(ResetEvent(io->overlapped.hEvent));
            if (io->status)
            {
                /* error return for a non-queued operation */
                SocketHandleSetLastError(sh, io->status);
                ret = -1;
                msg(D_WIN32_IO | M_ERRNO, "WIN32 I/O: Completion non-queued error");
            }
            else
            {
                /* successful return for a non-queued operation */
                if (buf)
                {
                    *buf = io->buf;
                }
                ret = io->size;
                dmsg(D_WIN32_IO, "WIN32 I/O: Completion non-queued success [%d]", ret);
            }
            break;
 
        case IOSTATE_INITIAL: /* were we called without proper queueing? */
            SocketHandleSetInvalError(sh);
            ret = -1;
            dmsg(D_WIN32_IO, "WIN32 I/O: Completion BAD STATE");
            break;
 
        default:
            ASSERT(0);
    }
 
    if (from && ret > 0 && sh.is_handle && sh.prepend_sa)
    {
        ret -= read_sockaddr_from_packet(buf, &from->dest.addr.sa);
    }
 
    if (!sh.is_handle && from)
    {
        read_sockaddr_from_overlapped(io, &from->dest.addr.sa, ret);
    }
 
    if (buf)
    {
        buf->len = ret;
    }
    return ret;
}
 
#endif /* _WIN32 */
 
/*
 * Socket event notification
 */
 
unsigned int
socket_set(struct link_socket *s, struct event_set *es, unsigned int rwflags, void *arg,
           unsigned int *persistent)
{
    if (s)
    {
        if ((rwflags & EVENT_READ) && !stream_buf_read_setup(s))
        {
            ASSERT(!persistent);
            rwflags &= ~EVENT_READ;
        }
 
#ifdef _WIN32
        if (rwflags & EVENT_READ)
        {
            socket_recv_queue(s, 0);
        }
#endif
 
        /* if persistent is defined, call event_ctl only if rwflags has changed since last call */
        if (!persistent || *persistent != rwflags)
        {
            event_ctl(es, socket_event_handle(s), rwflags, arg);
            if (persistent)
            {
                *persistent = rwflags;
            }
        }
 
        s->rwflags_debug = rwflags;
    }
    return rwflags;
}
 
#if UNIX_SOCK_SUPPORT
 
void
sd_close(socket_descriptor_t *sd)
{
    if (sd && socket_defined(*sd))
    {
        openvpn_close_socket(*sd);
        *sd = SOCKET_UNDEFINED;
    }
}
 
/*
 * code for unix domain sockets
 */
 
const char *
sockaddr_unix_name(const struct sockaddr_un *local, const char *null)
{
    if (local && local->sun_family == PF_UNIX)
    {
        return local->sun_path;
    }
    else
    {
        return null;
    }
}
 
socket_descriptor_t
create_socket_unix(void)
{
    socket_descriptor_t sd;
 
    if ((sd = socket(PF_UNIX, SOCK_STREAM, 0)) < 0)
    {
        msg(M_ERR, "Cannot create unix domain socket");
    }
 
    /* set socket file descriptor to not pass across execs, so that
     * scripts don't have access to it */
    set_cloexec(sd);
 
    return sd;
}
 
void
socket_bind_unix(socket_descriptor_t sd, struct sockaddr_un *local, const char *prefix)
{
    const mode_t orig_umask = umask(0);
 
    if (bind(sd, (struct sockaddr *)local, sizeof(struct sockaddr_un)))
    {
        msg(M_FATAL | M_ERRNO, "%s: Socket bind[%d] failed on unix domain socket %s", prefix,
            (int)sd, sockaddr_unix_name(local, "NULL"));
    }
 
    umask(orig_umask);
}
 
socket_descriptor_t
socket_accept_unix(socket_descriptor_t sd, struct sockaddr_un *remote)
{
    socklen_t remote_len = sizeof(struct sockaddr_un);
    socket_descriptor_t ret;
 
    CLEAR(*remote);
    ret = accept(sd, (struct sockaddr *)remote, &remote_len);
    if (ret >= 0)
    {
        /* set socket file descriptor to not pass across execs, so that
         * scripts don't have access to it */
        set_cloexec(ret);
    }
    return ret;
}
 
int
socket_connect_unix(socket_descriptor_t sd, struct sockaddr_un *remote)
{
    int status = connect(sd, (struct sockaddr *)remote, sizeof(struct sockaddr_un));
    if (status)
    {
        status = openvpn_errno();
    }
    return status;
}
 
void
sockaddr_unix_init(struct sockaddr_un *local, const char *path)
{
    local->sun_family = PF_UNIX;
    strlcpy(local->sun_path, path, sizeof(local->sun_path));
}
 
void
socket_delete_unix(const struct sockaddr_un *local)
{
    const char *name = sockaddr_unix_name(local, NULL);
    if (name && strlen(name))
    {
        unlink(name);
    }
}
 
bool
unix_socket_get_peer_uid_gid(const socket_descriptor_t sd, uid_t *uid, gid_t *gid)
{
#ifdef HAVE_GETPEEREID
    uid_t u;
    gid_t g;
    if (getpeereid(sd, &u, &g) == -1)
    {
        return false;
    }
    if (uid)
    {
        *uid = u;
    }
    if (gid)
    {
        *gid = g;
    }
    return true;
#elif defined(SO_PEERCRED)
    struct ucred peercred;
    socklen_t so_len = sizeof(peercred);
    if (getsockopt(sd, SOL_SOCKET, SO_PEERCRED, &peercred, &so_len) == -1)
    {
        return false;
    }
    if (uid)
    {
        *uid = peercred.uid;
    }
    if (gid)
    {
        *gid = peercred.gid;
    }
    return true;
#else  /* ifdef HAVE_GETPEEREID */
    return false;
#endif /* ifdef HAVE_GETPEEREID */
}
 
#endif /* if UNIX_SOCK_SUPPORT */